How the Translation Memory is Built
The TM is populated automatically whenever the AI translation engine saves translations. When an AI-generated translation is stored, Entri records the source string, the target string, and the language pair in the TM for future reference.TM entries are created when AI translations are saved (at Translated status). Human-edited translations are not automatically added to the TM at this time.
Organization-wide Sharing
The TM is scoped to your organization, not to a single project. A translation generated in your marketing site project is immediately available as a suggestion in your mobile app project or help center. This cross-project sharing ensures terminology stays consistent at scale.Fuzzy Matching
Not every new string will be an exact copy of something already in the TM. Entri’s fuzzy matching engine finds strings that are similar but not identical and presents them with a match percentage:| Match level | Description |
|---|---|
| 100% | Exact match — the source string is identical |
| 75–99% | High fuzzy match — one or two word differences |
| 50–74% | Moderate fuzzy match — useful as a starting point, needs review |
| Below 50% | Low similarity — not surfaced as suggestions |
Using TM Suggestions in the Editor
When you open a translation cell in the editor, Entri checks the TM in real time. If a match exists, a Memory badge may appear below the cell showing:- The matched source string and the percentage similarity
- The stored target translation
- Which project the match came from
AI Translation Integration
The TM works alongside AI Translation. When Entri generates an AI translation, it includes relevant TM matches in the prompt so the AI anchors its vocabulary to strings your team has already saved. This produces AI output that is more consistent with your existing translations.Searching and Browsing the TM
Navigate to Organization Settings → Translation Memory to search and manage your TM directly.- Full-text search — search by source or target string across all language pairs
- Language pair filter — narrow results to a specific source/target language combination
- Usage tracking — each TM entry shows
usageCount, which tracks how many times the same source text has been stored (e.g. duplicate strings across keys)
Best Practices
- Review AI translations regularly. The TM reflects what the AI has generated — reviewing and correcting translations keeps future suggestions accurate.
- Prune stale entries. If a source string changes significantly, delete the old TM entry to avoid misleading suggestions for future translators.